Low canopy temperature (CT) is an integrative trait “being associated with yield in a range of conditions”. It is indicative of the relative fitness of a genotype to the environment. Therefore, CT can be used as an indirect selection criterion for yield. Thermal measurements with handheld thermometers have their shortcomings, especially because genotypes should be measured at the same time to avoid distortions by changing environmental conditions. Main sources of short-term variability include wind, sunlight, clouds, and air temperature. Thus, thermal cameras mounted on drones are an interesting option to measure many plots in a relatively short time which reduces the variability between measurements. To get absolute CT measurements, calibrated thermal cameras must be used. Cooled thermal cameras are precise but heavy and cannot be mounted on a lightweight drone. For uncooled calibrated thermal cameras, a calibration must be done prior to use to get accurate absolute CT values. This takes specific system knowledge, and these cameras are still relatively heavy. To our knowledge, there are no off-the-shelf ready-to-use calibrated thermal cameras that could be operated on a drone without significant integration efforts. However, there are uncalibrated thermal cameras that can be operated with standard drones and standard software. This study seeked to improve the quality of CT measurements on wheat with an uncalibrated off-the-shelf thermal camera by analysing multiple georeferenced original images for each plot in a ray tracing approach. We exploited spatial information on plot location, temporal information on measurements and geometric information on the sun-plot-camera relations to correct for short-term-variability. Preliminary analysis shows that Cullis broad sense heritabilities of CT as high as 0.95 can be reached by this approach.
